run: introduce --remove-outs option

Signed-off-by: Ruslan Kuprieiev <[email protected]>

dvc/cli.py

@@ -285,6 +285,11 @@ def parse_args(argv=None):
                         help="Run this stage even if it has been already "
                         "ran with the same command/dependencies/outputs/etc "
                         "before.")
+    run_parser.add_argument(
+                        '--remove-outs',
+                        action='store_true',
+                        default=False,
+                        help="Remove outputs before running the command.")
     run_parser.add_argument(
                         'command',
                         nargs=argparse.REMAINDER,

dvc/command/run.py

@@ -31,7 +31,8 @@ class CmdRun(CmdBase):
                              cwd=self.args.cwd,
                              no_exec=self.args.no_exec,
                              overwrite=overwrite,
-                             ignore_build_cache=self.args.ignore_build_cache)
+                             ignore_build_cache=self.args.ignore_build_cache,
+                             remove_outs=self.args.remove_outs)
         except DvcException as ex:
             self.project.logger.error('Failed to run command', ex)
             return 1

dvc/project.py

@@ -364,7 +364,8 @@ class Project(object):
             cwd=os.curdir,
             no_exec=False,
             overwrite=False,
-            ignore_build_cache=False):
+            ignore_build_cache=False,
+            remove_outs=False):
         from dvc.stage import Stage
 
         with self.state:
@@ -377,7 +378,8 @@ class Project(object):
                                 metrics_no_cache=metrics_no_cache,
                                 deps=deps,
                                 overwrite=overwrite,
-                                ignore_build_cache=ignore_build_cache)
+                                ignore_build_cache=ignore_build_cache,
+                                remove_outs=remove_outs)
 
         if stage is None:
             return None

dvc/stage.py

@@ -340,7 +340,8 @@ class Stage(object):
               locked=False,
               add=False,
               overwrite=True,
-              ignore_build_cache=False):
+              ignore_build_cache=False,
+              remove_outs=False):
 
         stage = Stage(project=project,
                       cwd=cwd,
@@ -368,6 +369,10 @@ class Stage(object):
         stage.cwd = cwd
         stage.path = path
 
+        # NOTE: remove outs before we check build cache
+        if remove_outs:
+            stage.remove_outs(ignore_remove=False)
+
         if os.path.exists(path):
             if not ignore_build_cache and stage.is_cached():
                 Logger.info('Stage is cached, skipping.')

tests/test_run.py

@@ -185,6 +185,23 @@ class TestCmdRun(TestDvc):
         self.assertEqual(ret, 252)
 
 
+class TestRunRemoveOuts(TestDvc):
+    def test(self):
+        with open(self.CODE, 'w+') as fobj:
+            fobj.write("import sys\n")
+            fobj.write("import os\n")
+            fobj.write("if os.path.exists(sys.argv[1]):\n")
+            fobj.write("    sys.exit(1)\n")
+            fobj.write("open(sys.argv[1], 'w+').close()\n")
+
+        ret = main(['run',
+                    '--remove-outs',
+                    '-d', self.CODE,
+                    '-o', self.FOO,
+                    'python', self.CODE, self.FOO])
+        self.assertEqual(ret, 0)
+
+
 class TestCmdRunOverwrite(TestDvc):
     def test(self):
         ret = main(['run',